$600 a month. Five businesses. Zero full-time employees. One voice agent making 100 to 200 B2B cold calls a day in English, Ukrainian, and Russian. That's the stack I run from a laptop in Brooklyn. I'm going to tell you about the voice agent, because it's the piece nobody believes until they hear it pick up a phone.
Quick context so you know who's talking. I landed in Brooklyn in March 2022 with a wife, a two-year-old son being treated for brain cancer, one suitcase, and a laptop. We'd been in Kyiv ten days earlier. I spent the first months in the US figuring out how to feed my family. No CS degree. No engineering background. B2B distribution operator before the war, builder after. That's the whole resume.
Four years later my son is six, healthy, starting first grade, and the voice agent I built from that kitchen table is running outbound for a Ukrainian stone import business into the US market.
The business is OD Granite. Kitchen countertops, bathroom slabs, commercial jobs, sold B2B into the US market. Every lead is a phone call with a fabricator, a contractor, or a designer. The kind of outbound work you normally staff with 2 or 3 SDRs at $50 to $80K a year each in NYC. Call it $10 to $15K a month in payroll before you count benefits, desks, onboarding, ramp time, and the six weeks of dead air when one of them quits.
I don't pay that. I pay for an agent.
The agent dials prospects all day. It opens with a real reason for the call, asks two or three qualifying questions, handles the normal objections ("send me an email," "we already have a supplier," "who is this again"), and when someone is interested it drops a meeting straight into my calendar. If the person on the other end answers in Ukrainian or Russian instead of English, it switches language mid-sentence and keeps going. That one detail kills every off-the-shelf tool I looked at. The US construction supply chain is full of Ukrainian and Eastern European operators. An English-only bot dies at hello on half my list.
Volume sits between 100 and 200 dials a day depending on list quality. Out of that I get 6 to 10 qualified meetings on my calendar per week. Not leads. Not MQLs. Actual meetings with a human who picked up a phone, had a real conversation, and agreed to a follow-up slot.
That's the conservative floor I'm willing to defend. I had the number higher in earlier drafts. I walked it down because I only put numbers on this website that I can pull up the logs for if somebody asks me at a dinner. The actual ceiling is higher. The floor is what I'll stand behind on the record.
This one works. Plenty of the others didn't, and I want to be honest about that because anyone who tells you their AI stack is a straight line is lying or selling a course.
I've killed agents that cost me real money. One was an outbound email system I babied for weeks, burning API credits every night on messages nobody wanted to read. I kept tweaking the prompt, kept telling myself the next iteration would hit, kept sending. It didn't hit. I shut it off and ate the cost. Another was a qualification bot that scored leads in a way that looked smart in the dashboard and stupid in practice. It was promoting tire-kickers and burying the real buyers. I killed it and went back to a simple filter that did better.
My rule now is brutal and boring. I build something. I watch the one number I care about for two weeks. If the number isn't moving, the agent is dead. I've killed more agents than I've kept. The voice agent is the one that survived, and the reason I trust it is because everything around it got buried.
The voice agent breaks about once a week. Usually something small. A phone provider hiccup. A prompt that drifts after I tune it. A schedule edge case where the agent tries to call a contractor at 4 a.m. their time. I fix most of it from my phone in a soccer parking lot while my kid is at practice. That's the whole point of the stack. It runs without me most of the time, and when it breaks I don't need to be at a desk.
The agent doesn't close deals. It books meetings. I still have to take the calls, answer the hard questions, negotiate the prices, walk through the stone options, handle the "my last supplier screwed me on lead time" stories. The last 20% of the sale is a human job. It always will be. But the first 80%, the part that used to eat a whole team's week, runs on its own.
And it's not the only one. The same stack drafts every response to 700+ reviews on my Amazon brand at 4.6 stars with zero customer service headcount. It delivers a daily profit and loss across all five businesses to my phone at 9 a.m. in thirty seconds. It watches ad spend across Meta and Amazon PPC and flags overspend before it hits the daily cap. Forty-plus automated social posts a week across five brands, cron-verifiable. The voice agent is the loudest piece because it talks. The rest carries the same weight, it's quieter.
If you're running a business in the $1M to $10M range and you're about to hire your next SDR, your next researcher, your next "someone to handle outbound," stop for a second and do the math. Cold dialing. Qualifying. First-touch. Follow-up. Meeting booking. Every piece of that job is either automatable or it isn't. The automatable pieces should run on a $600 a month stack, not a $10K a month headcount.
I'm not saying fire your team. I'm saying don't hire the next one before you've tried to build it. Most of the time you'll fail. A few times it'll work and you'll wonder why you waited. And once in a while you'll build something that replaces 2 to 3 people in a payroll line, and you'll spend the rest of the year asking yourself what else you've been paying for out of habit.
The guys who already figured this out aren't winning because they hired better people. They're winning because they stopped hiring for jobs that don't need a person anymore. It's a quiet advantage and it compounds every month you ignore it.
If you want a second set of eyes on your own operation, that's what I do now. I sit with founders in the $1M to $10M range, map the jobs their team is doing, and tell them which ones should be running on an agent by next quarter. No slide decks. No buzzword bingo. A real audit from someone who runs the stack on his own money first.
Book a call on the button below. Worst case you get an honest opinion. Best case I save you the next SDR hire.
It takes a few months of looking stupid first. I looked stupid for a year. The voice agent is what came out the other end.
Related resources:
Want to build AI systems that work?
Book a Free Strategy Call